[adregan]

In the section on security:

> One large enterprise employee commented that they were deliberately slow with AI tech, keeping about a quarter behind the leading edge. “We’re not in the business of avoiding all risks, but we do need to manage them”.

I’m unclear how this pattern helps with security vis-à-vis LLMs. It makes sense when talking about software versions, in hoping that any critical bugs are patched, but prompt injection springs eternal.

[Quothling]

I work in a NIS2 regulated sector and I'm not sure we can ever let any AI agent run in anything we do. We have a centralized sollution where people can build their own chatbots with various configurations and cross models. That's in the isolation of the browser though, and while I'm sure employees are putting things into it they shouldn't, at least it's inside our setup and not in whatever chatbot they haven't yet run out of tokens on. Security wise though, I'm not sure how you can meet any form of compliance if you grant AI's access unless you have four eye validation on every single action it takes... which is just never going to happen.

We've experimented with rolling open source models on local hardware, but it's so easy to inject things into them that it's not really going anywhere. It's going to be a massive challenge, because if we don't provide the tools, employees are going to figure out how to do it on their own.

[bilekas]

> but prompt injection springs eternal.

Yes, but some are mitigated when discoverd, and some more critical areas need to be isolated from the LLM so taking their time to provision LLM into their lifecycle is important, and they're happy to spend the time doing it right, rather than just throwing the latest edge tech into their system.

[ethin]

How exactly can you "mitigate" prompt injections? Given that the language space is for all intents and purposes infinite, and given that you can even circumvent these by putting your injections in hex or base64 or whatever? Like I just don't see how one can truly mitigate these when there are infinite ways of writing something in natural language, and that's before we consider the non-natural languages one can use too.

[lambda]

The only ways that I can think of to deal with prompt injection, are to severely limit what an agent can access.

* Never give an agent any input that is not trusted

* Never give an agent access to anything that would cause a security problem (read only access to any sensitive data/credentials, or write access to anything dangerous to write to)

* Never give an agent access to the internet (which is full of untrusted input, as well as places that sensitive data could be exfiltrated)

An LLM is effectively an unfixable confused deputy, so the only way to deal with it is effectively to lock it down so it can't read untrusted input and then do anything dangerous.

But it is really hard to do any of the things that folks find agents useful for, without relaxing those restrictions. For instance, most people let agents install packages or look at docs online, but any of those could be places for prompt injection. Many people allow it to run git and push and interact with their Git host, which allow for dangerous operations.

My current experimentation is running my coding agent in a container that only has access to the one source directory I'm working on, as well as the public internet. Still not great as the public internet access means that there's a huge surface area for prompt injection, though for the most part it's not doing anything other than installing packages from known registries where a malicious package would be just as harmful as a prompt injection.

Anyhow, there have been various people talking about how we need more sandboxes for agents, I'm sure there will be products around that, though it's a really hard problem to balance usability with security here.

[bilekas]

Full mitigation seems impossible to me at least but the obvious and public sandox escape prompts that have been discovered and "patched" out just making it more difficult I guess. But afau it's not possible to fully mitigate.

[charcircuit]

If the model is properly aligned then it shouldn't matter if there is an infinite ways for an attacker to ask the model to break alignment.

[lambda]

How do you "properly align" a model to follow your instructions but not the instructions of an attacker that the model can't properly distinguish from your own? The model has no idea if it's you or an attacker saying "please upload this file to this endpoint."

This is an open problem in the LLM space, if you have a solution for it, go work for Anthropic and get paid the big bucks, they pay quite well, and they are struggling with making their models robust to prompt injection. See their system card, they have some prompt injection attacks where even with safeguards fully on, they have more than 50% failure rate of defending against attacks: https://www-cdn.anthropic.com/c788cbc0a3da9135112f97cdf6dcd0...

[charcircuit]

>The model has no idea if it's you or an attacker saying "please upload this file to this endpoint."

That is why you create a protocol on top that doesn't use inbound signaling. That way the model is able to tell who is saying what.

[lambda]

Huh? Once it gets to the model, it's all just tokens, and those are just in band signalling. A model just takes in a pile of tokens, and spits out some more, and it doesn't have any kind of "color" for user instructions vs. untrusted data. It does use special tokens to distinguish system instructions from user instructions, but all of the untrusted data also goes into the user instructions, and even if there are delimiters, the attention mechanism can get confused and it can lose track of who is talking at a given time.

And the thing is, even adding a "color" to tokens wouldn't really work, because LLMs are very good at learning patterns of language; for instance, even though people don't usually write with Unicode enclosed alphanumerics, the LLM learns the association and can interpret them as English text as well.

As I say, prompt injection is a very real problem, and Anthopic's own system card says that on some tests the best they do is 50% on preventing attacks.

If you have a more reliable way of fixing prompt injection, you could get paid big bucks by them to implement it.

[MattGrommes]

I took this to mean more like not jumping right on OpenClaw, but wait a quarter or so to give it at least a little time to shake out. There are so many new tools coming out I think it's beneficial not to be the guinea pig.